fiscal multipliers in a panel of countries - bank of · pdf filefiscal multipliers in a panel...
TRANSCRIPT
Banco de México
Documentos de Investigación
Banco de México
Working Papers
N° 2014-15
Fiscal Mult ipl iers in a Panel of Countries
July 2014
La serie de Documentos de Investigación del Banco de México divulga resultados preliminares de
trabajos de investigación económica realizados en el Banco de México con la finalidad de propiciar elintercambio y debate de ideas. El contenido de los Documentos de Investigación, así como lasconclusiones que de ellos se derivan, son responsabilidad exclusiva de los autores y no reflejannecesariamente las del Banco de México.
The Working Papers series of Banco de México disseminates preliminary results of economicresearch conducted at Banco de México in order to promote the exchange and debate of ideas. Theviews and conclusions presented in the Working Papers are exclusively the responsibility of the authorsand do not necessarily reflect those of Banco de México.
Juan ContrerasBanco de México
Holly Bat te l leWeSpire
Fiscal Mult ipl iers in a Panel of Countr ies*
Abstract: We estimate fiscal multipliers in a panel of countries using dynamic panel techniques andquarterly data for 55 countries. By using a GMM estimator and lagged dependent variables asinstruments in a SVAR model, we attempt to correct for the biases present in this setting, to alleviateconcerns about causality, and to decrease potential effects of third factors. Contrary to previous research,we find no strong evidence of monetary accommodation, a positive and larger fiscal multiplier indeveloping than in high-income countries, and zero in high-debt countries and in flexible exchange ratescountries.Keywords: Fiscal multipliers, Panel of countries, SVAR, GMM.JEL Classification: E62, E63, H60.
Resumen: En este trabajo estimamos los multiplicadores fiscales para un panel de 55 países usandotécnicas de panel dinámico y datos trimestrales. Al usar un estimador GMM y rezagos de las variablesdependientes como instrumentos, intentamos corregir los sesgos presentes en este caso, además dealiviar los problemas de causalidad y de disminuir los efectos potenciales de otros factores. Al contrariode las conclusiones reportadas en la literatura, encontramos que los multiplicadores fiscales sonpositivos en los países en desarrollo y mayores que los estimados para los países desarrollados. En lamisma línea, encontramos además que estos multiplicadores son cero en los países con deuda alta y enlos países con tipo de cambio flexible. Tampoco encontramos evidencia importante que sugieraacomodación monetaria.Palabras Clave: Multiplicadores fiscales, Panel de países, SVAR, GMM.
Documento de Investigación2014-15
Working Paper2014-15
Juan Cont re ras y
Banco de MéxicoHol ly Ba t te l l e z
WeSpire
*We thank Nicolás Amoroso, Wendy Edelberg, Jonathan Huntley, Bill Randolph, Mark Lasky, and seminarparticipants at the Congressional Budget Office, the Bank of México, the Southern Economics Association 2011meetings and the Latin American Econometric Society 2013 meetings for comments that helped to improve thispaper. The views expressed in this paper are the authors' and should not be interpreted as Bank of México's. y Dirección General de Investigación Económica. Email: [email protected]. z WeSpire. Email: [email protected].
1 Introduction
The size of fiscal multipliers, or the change in output in response to a change in fiscal
policy, remains a source of disagreement among economists despite the importance to
public policy. This disagreement comes mainly from the differences in the methodologies
and data used by different researchers to avoid the potential bias caused by endogeneity
between output (GDP) and fiscal policy. A growing economy may be responsible for
increases in government spending, but observed increases in government spending may
cause a growing economy. In addition, a third fact may cause changes in both government
spending and output. For example, a sudden discovery of a natural resource may trigger
at the same time an increase in output and an increase in government spending. Fiscal
policy is also known to be implemented with a lag, which is tightly linked to identifying
the anticipation from the part of private agents of changes in fiscal policy that may affect
their behavior.
This paper contributes to the literature of fiscal multipliers by showing, contrary to
previous research, that there is no strong evidence of monetary accommodation,1 that
fiscal multipliers are positive and larger in developing than in high-income countries,
and that they are zero in high-debt countries and in flexible exchange rates countries.
Other results show fiscal multipliers that are positive and statistically different from
zero, with an impact multiplier of 0.3 and a long run multiplier between 0.9 and 1.0. We
also find that not controlling for the interest rates (and implicitly for monetary policy)
or for the real exchange rate, makes the estimates smaller. In our estimations, private
consumption response is positive to fiscal shocks at different horizons.
More generally, results in this paper question the robustness of the conclusions drawn
in previous VAR literature with panel data. By using dynamic panel data techniques -a
Generalized Method of Moments estimator that instruments the endogenous variables
as Holtz-Eakin, Newey and Rosen (1988)- instead of using an OLS estimator with fixed
effects as is common in the literature, we attempt to further correct for the potential
1We refer to monetary accommodation to the active role that the monetary authority can play toenhance the effects of fiscal policy.
1
biases induced by the correlation of the lags with the error terms known to be present
in this type of setting. Because this technique uses instruments for the endogenous vari-
ables, it also ameliorates two additional sources of endogeneity, namely the simultaneity
between government spending and output growth and the likely presence of a third fac-
tor that may affect both government spending and output. In fact, OLS overestimates
the fiscal multipliers, as is shown in section 5 of the paper. On the other hand, by using
a more comprehensive dataset than previous studies, we are able to establish that the
sample selection is important when estimating fiscal multipliers in a panel of countries.
This result is not specific to our dataset as we also show later in the paper.
In the next section we discuss the potential issues in estimating fiscal multipliers. In
section 3 we present our identification strategy. We describe our data in section 4 and
show our results in sections 5. In this section we also analyze the sources of differences
between our results and previous research. We conclude in section 6.
2 Issues in Estimating Fiscal Multipliers
Studies use mainly two alternative approaches and identifying assumptions to solve the
endogeneity problem.2,3 The first approach, the “narrative approach”, uses information
about shocks that are unexpected and independent of the state of the economy, and
which prompt the government to spend more. Using such strategy, Ramey and Shapiro
(1997) create a univariate autoregressive model and use it to estimate the effects that
military buildups have on a variety of macroeconomic variables. In their study, the
military buildup is signaled by a dummy variable to indicate the Korean War, the Viet-
nam War and the Carter-Reagan buildup. Ramey and Shapiro explain that military
buildups are natural shocks to the economy because they usually occur rapidly and un-
expectedly. Moreover, the military buildup variable is attractive because it is likely to
2Previous research usually ignores the endogeneity problem that refers to a third factor as the causeof both the increase in government spending and the change in output.
3A different approach is taken by Suarez-Serrato and Wingender (2011), who use the fact that alarge number of federal spending and transfer programs depend on population estimates, which changeduring Census years due to a change in methodology. They find a multiplier of about 1.9.
2
be exogenous to other macroeconomic variables, allowing them to analyze a pure shock
to GDP. They find that military buildups have a positive impact on GDP for three
years and reach a peak impact of 3 percent after an the onset of one of those episodes,
which correspond to an increase of around 1% in government spending according to
Edelberg et al, (1997).4 The main critique of early implementations of this approach
is the limited amount of episodes available to identify the fiscal multipliers, although
this issue is corrected in Ramey (2011).5 In this later study, Ramey (2011) expanded
on her previous work to include two new variables that measure military buildup an-
ticipations and to use a SVAR approach instead of a univariate approach. Her findings
indicate that government spending multipliers range from 0.6 to 1.1. She finds negative
private consumption responses to government spending and shows that previous findings
of positive consumption responses, as in Blanchard and Perotti (2002), come from the
fact that agents anticipate the government spending more than one quarter in advance,
invalidating the identification assumption that within a quarter shocks to output do not
cannot affect government spending. Other authors have used variations of the strategy
of identifying fiscal multipliers using military spending. For example, Nakamura and
Steinsson (2012) used historical data on military procurement across US states to esti-
mate a fiscal multiplier of about 1.5, and Barro and Redlick (2011) use military spending
to estimate a multiplier that is below 1.
The second approach uses a structural vector auto regression (SVAR) to identify the
effects of spending on output by assuming that within a quarter the government cannot
respond with fiscal policy to unexpected changes in output, and that either spending
doesn’t respond to taxes or vice versa within that quarter. Blanchard and Perotti (2002)
study the effects of shocks to government spending and taxes on United States economic
4Edelberg, Eichenbaum and Fisher (1997) expand on an earlier version of the Ramey and Shapiro(1997) study, and use a vector autoregressive model to analyze the effect of an exogenous shock to USgovernment purchases on various macroeconomic variables. Edelberg, Eichenbaum and Fisher test theuncertainty surrounding the dates used for the three military buildups in the Ramey and Shapiro paper,finding that the dates chosen are robust. Additionally they find that an exogenous shock to governmentpurchases has a similar effect on GDP as the military buildup shocks in the Ramey and Shapiro paper.
5Another critique of this approach poses that wars could not be entirely exogenous, and insteadpolitically motivated to increase output. This is the same reverse causality mechanism mentionedbefore in the introduction.
3
activity using this identification strategy. By using a mixed structural VAR/event study
approach, their results suggest that positive shocks to government spending have a
positive effect on output. Specifically, they find that GDP increases on impact by 0.84
dollars following a positive government spending shock, then declines, and rises again,
to reach a peak effect of 1.29 dollars per 1 dollar of spending after almost 4 years. Other
studies for the US, which also use a SVAR, include Mountford and Uhlig (2009), who
find an impact multiplier of 0.65 and a long run multiplier of -1, Fatas and Mihov (2001),
who find a long run multiplier similar to the estimates by Blanchard and Perotti (2002),
and Auerbach and Gorodnichenko (2012), who use semiannual data to estimate fiscal
multipliers through the business cycle.6
The SVAR approach has been used by studies that use a panel of countries. Per-
otti (2004) analyzes the effect of fiscal policy shocks on the economies of five separate
OECD countries (the United States, West Germany, the United Kingdom, Canada, and
Australia) using a SVAR with a large dataset that begins in 1960 and terminates in
2001. He breaks his sample into pre-1980 and post-1980, finding that the effects of fiscal
policy tend to be smaller than other studies suggest: a government spending multiplier
greater than 1 is only estimated in the United States prior to 1980. In the post-1980
sample, Perotti estimates a GDP cumulative response to a spending shock to range
from anywhere between negative 2.25 to positive 0.77 percent. Beetsma and Giuliodori
(2011) find a multiplier of 1.6 for the European countries, although using annual data.
More recently, Ravn et al. (2012), use a SVAR from four industrialized countries and
document a positive fiscal multiplier, a positive response of private consumption, and a
depreciation of the real exchange rate.
Ilzetzki, Mendoza and Vegh (IMV) (2013) is the closest work related to this paper.
They use a SVAR model with Blanchard and Perotti (2002) identification strategy to
analyze the impact of government expenditure shocks on output for 44 countries at a
quarterly frequency. Overall, they conclude that fiscal multipliers are much smaller than
other studies suggest. Additionally, their results suggest that country characteristics are
6A closely related literature investigates the effect of tax policy on output in a SVAR context, sharingsimilar identification challenges. See for example Mertens and Ravn (2012).
4
crucial in determining the size of the multiplier, finding: an increase in government con-
sumption leads to a higher output effect in industrial countries compared to developing
countries; the fiscal multiplier is relatively large in economies operating under predeter-
mined exchange rates, but zero in economies operating under flexible exchange rates;
open economies have smaller multipliers than closed economies; and fiscal multipliers
are negative in high-debt countries. We depart from their study in two important ways.
First, we use a different sample with more countries (55) over a longer period (1988
to 2010), having available more than 3000 observations. Second, we use panel SVAR
estimator that corrects for the correlation between the error terms and the explanatory
variables present in this type of setting. This estimator uses lagged values of the en-
dogenous variables as instruments, which ameliorates concerns about the possibility that
fiscal policy is anticipated by economic agents, and about the possibility that a third
unobserved factor drives the results. We obtain different results than they do and test
the sources of differences. In particular, we do not get negative multipliers for develop-
ing countries and we do not see any differences in the output response to government
spending between high and low debt countries. Overall, although we can replicate their
results when we use their method and sample of countries, we obtain different results
when we use their dataset but with a different sample of countries.
3 Identification
We use the basic identification strategy of Blanchard and Perotti (2002) plus a correction
for the endogeneity present in the panel SVAR using a generalized method of moments
estimator and lagged endogenous variables as instruments. By accounting for the dy-
namic correlation between the lags of the variables with the error terms, we are able to
ameliorate the potential biases that come from anticipation effects and from the presence
of a third factor that causes movements in both government spending and output. The
panel SVAR is an extension of the structural VAR and allows for unobserved individ-
ual heterogeneity in each country characteristics through fixed effects. We estimate the
5
following equation:
zi,t = βi,t + β(L)zi,t−1 + εi,t (1)
where βi,t is a vector that includes fixed effects and a common quadratic trend. zi,t
is a vector [G, Y, r, x, Z], where G is log of per-capita government spending, Y is log of
per capita output, r is the policy interest rate, x is an index of the real effective interest
rate, and Z is a set of variables that includes the log of total employment and the log
of per capita consumption. The identification strategy treats the shocks to government
spending as exogenous to output within a quarter. We analyze the response the variables
in z have to a shock in government spending. We take 4 lags as our benchmark, but
results are robust to the consideration of a structure with 8 lags. We detrended the
data using a quadratic trend, but results are almost identical with a linear trend. Our
benchmark estimation includes the interest rate and the real exchange rate, but we also
report some robustness results if we do not include them in the estimation.
We are inerested in the second equation of this system, and in particular our focus
is the effect of government spending G over output Y. The response of the interest rate
to changes in government spending could also shed light on the monetary authorities
responses to monetary policy, an issue we will explore later.
There are two potential sources of biases in this type of estimation that come from
violating the conditional independence assumption (unobserved state variables follow an
i.i.d. process and are conditional independent of observed state variables). First, a third
unobserved factor can affect at the same time government spending and output; an ex-
ample could be some political considerations lead to war and then to higher government
spending and output. And second, fiscal policy can be anticipated with more than one
quarter so that the shocks to fiscal policy and the shocks to output are correlated.
We control for the bias sources in two ways: by using the panel VAR estimator, which
instruments the endogenous variables with lagged values, and by using fixed effects in the
error term. With respect to the possible presence of a third unobserved factor, lagged
6
endogenous variables as instruments help to control for underlying time-variant third
factors, while fixed effects control for this bias if the third factor is fixed through time.
Fixed effects also control for variation in characteristics across countries and for the
presence of individual unobserved heterogeneity. The problem with fixed effects is that
they are correlated with the lags of the dependent variables. Fixed effects are usually
removed by taking first differences, but in this case first differences will yield again
biased coefficients because the differenced variables are correlated with the differenced
error terms. The method we apply, and that was proposed by Holtz-Eakin et al. (1988),
is a generalized method of moments estimator that uses lagged instruments to overcome
this endogeneity biases. To calculate the standard errors of the impulse-responses, we
follow a bootstrapping procedure in which we generate random draws of the coefficients
and calculate for each draw the impulse-response. We repeat this procedure 500 times.7
The estimator we used also helps to control for the second bias source, so that shocks
to output and shocks to government spending are not correlated. As mentioned above,
using this estimator we are able to account for the correlation of the lagged variables
with the error term, correcting for the potential bias induced by the anticipation effects
that this error term contains.8 In addition, Judson and Owen (1999) conclude that for
practical purposes, the type of GMM estimator that we use in this paper is the best
option to estimate the parameters in unbalanced panels with a small time dimension as
is the case in our dataset.
4 Data
We compiled a quarterly panel dataset that begins in 1988 and goes through the fourth
quarter of 2010. We identify a total of 55 countries including countries in Latin America,
7We use the codes developed by Inessa Love (2006) as the base for our estimations.8In a previous version of this paper, we also used the narrative approach to identify the effects of
government spending on output through wars as exogenous shocks. With different data sources, weidentified 22 war episodes. The estimates we obtained using this method had large standard errors,possibly because of the small number of war episodes, and we could not get statistical identification ofthe effects.
7
Asia/Pacific and the OECD countries. Table 1 shows the countries included in the
sample.9 We included data from the last recession (2008 to 2010) in the estimation, but
results were almost unchanged when we excluded this period.
For the macroeconomic series, including total government consumption, Gross Do-
mestic Product (GDP), private consumption, civilian employment,10 and GDP price
deflators, we use quarterly series that are not interpolated but directly reported from
central banks, governments and statistical offices. We take the data from Haver An-
alytics, a private company that sells data services. All series are adjusted seasonally
and we deflate the data to real terms using each country’s GDP deflator. Countries
reported data in thousands, millions, billions or trillions, so all data were also converted
into millions. We use quarterly policy interest rate (discount rate) from the IMF and
monthly policy rates from the sources shown for each country in case they do not ap-
pear in the IMF dataset. In the last case, we collapsed all monthly data to a quarterly
frequency. We use the index of the real effective exchange rate, Wholesale Price Index
as reported by the IMF and broad indices of real exchange rates reported by the Bank
of International Settlements. Table 2 shows basic statistics for the variables of interest.
To help control for differences between countries, we converted all macro data into per
capita terms, following Ramey (2012), by dividing the data by each country’s population.
We obtain the population data from the World Bank’s World Development Indicators
2012 dataset. This data is annual, so we interpolated it into quarterly data. Results are
robust when we consider aggregate series instead of per capita variables.
We follow IMV method to classify the exchange rate regimes, which is based on the
de-facto classification of Ilzezki, Reinhart and Rogoff (2008).11 A list of the exchange
9The complete list of countries and periods for each series is listed in a separate supplement accom-panying this document.
10Countries where civilian employment is not available include (parentheses indicate the proxy usedin different controls): Brazil (economically active population), Ecuador (employment: global occupa-tion rate), Pakistan (total employment), Peru (employed in metropolitan Lima), and the Philippines(employment).
11A country is considered to have a fixed exchange rate if during 8 quarters or more it has no legaltender, hard pegs, crawling pegs, and de facto or pre-announced bands or crawling bands with marginsof no larger than +/- 2 percent. All other episodes are considered flexible.
8
rate regimes for the countries is shown in Table 3. We also follow IMV to classify a
country’s openness by a de-facto measure of a country’s trade ratio (defined as exports
plus imports to GDP). If a country’s ratio is greater than 60 percent it is considered open
(see Table 4 for a complete list of open and closed economies). We use the Reinhart and
Rogoff (2010) database of national debt to classify countries according to their ratios of
general government debt to GDP (see Table 5). We use a threshold of 60 percent above
which a country is classified as high-debt country. And, finally, we use the 2010 World
Bank classification of developing vs. high income countries.
Although the data set used by IMV is similar to the one used in this paper, there are
important differences between them. First, they use 44 countries and we use 55.12 The
second difference is that, while we have the same time period for all the sample, countries
in the IMV dataset have different time periods. Those features make our dataset 50%
larger than the IMV dataset: our total number of observations in the pooled data is about
3900 while the IMV dataset has around 2500.13 With respect to the quality of our data,
we rely on the reported values from central banks and statistical agencies. Although it
is true that some countries do not have over all the period the same methodology to
collect and report data as is the case in the IMV dataset, we obtain similar results to
IMV when we use the same sample of countries, suggesting data quality is comparable.
5 Results
Figure 1 shows statistically and significant positive multiplier estimates through all the
horizon of analysis.14 The impact multiplier is 0.36 and the long run multiplier is 0.92.
12The countries that appear in the IMV dataset but do not appear in our dataset are Botswana,Bulgaria, Croatia, Estonia, Israel, Latvia, Lithuania, Romania, Slovenia and Uruguay. It can be seenthat they are predominantly former Soviet Union countries (7 out of 10). At the same time, thecountries that appear in our data set but do not appear in the IMV dataset are Austria, Bolivia,China, Costa Rica, Guatemala, Hong Kong, India, Indonesia, Japan, Korea, Luxembourg, New Zealand,Pakistan, Paraguay, Philippines, the Russian Federation, Singapore, Switzerland, Taiwan, Venezuelaand Vietnam.
13The number of observations that we use in the main estimation is 3131. If we consider the IMVdata, only 2241 can be used with our estimator.
14We calculate two types of multipliers: the impact multiplier, which measures the change in out-put in response to a one unit change in government spending in a given quarter, and the cumulative
9
Those results hold whether we include or not the interest rate and the real exchange
rate, although the numbers show a small change. In principle, controlling for the interest
rate should isolate the effect of monetary and fiscal policy, but the small change in the
estimates for the multipliers suggest that on average the monetary authority does not
accommodate much to fiscal policy. To illustrate this point, Figure 2 shows the response
of output, interest rate and real exchange rate to a one unit standard deviation shock in
government spending.15 Although the interest rate decreases in the first quarter, it in-
creases afterwards and is not statistically different to zero, suggesting no accommodation
from the monetary authority after the first quarter of the shock. At the same time, the
real exchange rate depreciates in the first quarters, but its change is not different from
zero from the third quarter on of the shock. This suggests a non persistent effect in the
real exchange rate of higher demand and imports. Ravn et. al. (2012) also observe a de-
preciation of the exchange rate in response to an increase in government purchases, and
propose an explanation based on deep habits, who cause markups to decline in markets
with strong aggregate demand, such that when government spending increases, markups
fall on domestically sold goods, depreciating the exchange rate. Although those numbers
are not too big, they imply a positive response of output to government spending, and
may imply larger multipliers during some periods of the business cycles.
Positive multipliers are consistent in principle with both the Keynessian and neoclas-
sical models.16 In the neoclassical model (see for example Baxter and King (1993)), an
increase in government spending creates a negative wealth effect for households because
it has to be matched by an increase of taxes in the future. Individuals reduce con-
sumption and leisure because of the negative wealth effect, increasing at the same time
labor supply and driving down the wage rate. Higher labor supply, in turn, increases
multiplier, which measures the cumulative change in output divided by the cumulative change in gov-ernment spending over a determined time horizon. Because all our specifications are in logs, we haveto multiply those values by the ratio of average output to average government spending to obtain thefiscal multipliers.
15The impulse response function of the variable X with respect to the variable Y is the response ofthe variable X to a one unit standard deviation shock in the variable Y. The units are in percentagepoints. The confidence intervals represent 90% of the distribution (5% the lower and 95% the higher).
16Positive multipliers also suggest negative output effects of fiscal consolidations. Under specificcircumstances, those effects may occur (Giavazzi and Pagano (1990), Alesina and Ardagana (1998)),but on average for our sample, results suggest that is not the case.
10
output. The main difference between the neoclassical model and the new Keynesian
model comes from the response of private consumption, which decreases in the neoclas-
sical model but increases in the new Keynessian model. Consumption may increase in
the new Keynesian model when government spending increases because nonseparability
between consumption and leisure, because the aggregate demand for labor shifts with
counter-cyclical markups, because nominal rigidities, because increasing returns in pro-
duction or because rule of thumb consumers. The key issue is that consumption increases
when the real wage does not change or when it increases. This effect can be attained
when the labor demand curve shifts outwards, at the same time that the labor supply
shifts outwards, such that the real wage does not fall. The fact that the multipliers are
on average lower than 1 may suggest some crowding out because output rises less than
government spending. Another point to notice is that different sources of government
financing might have a different effect in the short and in the long run. Higher debt, for
example, could affect long-run sustainability and thus current fiscal policy as well.
We find that private consumption responds positively to government spending in our
panel of countries, giving support to the Keynesian theories (see figures 3 and 4).17
Although always positive, this response in statistically significant only until the third
year, suggesting that probably the effects on consumption dilute with time. Consistent
with theory, we find that employment increases in all cases. This result is consistent
with the result that Blanchard and Perotti (2002) find for the United States, but is
the opposite of the results of Ramey and Shapiro (1997), Edelberg et al (1999) and
Ramey (2009). Ramey argues that the differences come from anticipation effects: once
she controls for expectations using the identification strategy of Blanchard and Perotti
(2002), private consumption actually falls. Our results are not particular to one country
or to a small number of episodes: we use a large sample of countries and correct for
additional potential biases coming from third factors affecting output and government
spending.
17Non-keynesian effects of fiscal policy, however, are not ruled out because of this result. Those effectsmay come, for example, through changes in expectations, credibility and interest rate premiums andlack of wealth effects on labor supply (Barry and Devereux, 2003). In fact, for our sample as shownlater, high-debt countries have a lower fiscal multipliers; this may come from of credibility concerns.
11
We next analyze the fiscal multipliers for countries operating in flexible vs. fixed
exchange rate regimes. When we do not control for the interest rate or the real exchange
rate, the impact and cumulative multipliers are statistically different throughout all
the time of analysis (figure 5). The impact multipliers are 0.57 and 0.27 in the case
of the fixed and flexible exchange rates, respectively, and the difference is statistically
significant. In a 5 year horizon, the cumulative multipliers are 1.58 and -0.27 respectively.
However, in the case of countries operating under flexible exchange rates, the multiplier
is not statistically different from zero except during the first year, a period in which it is
positive. IMV obtain the same relative results, although they get a negative cumulative
multiplier in the case of the flexible exchange rate regimes, and we do not. It is interesting
to note that once we control for the interest rate and the real exchange rate (figure 6), the
statistical differences disappear between multipliers. This shows that their differences
do come from the behavior of the real exchange rate and the interest rate. At the same
time, the fiscal multipliers for the countries operating under fixed exchange rates are
always statistically different from zero and positive, while they are not different from
zero for countries operating under flexible exchange rates.
Those results are in principle consistent with the Mundell Fleming model. Under
flexible exchange rate regimes, an increase in government spending causes an apprecia-
tion of the real exchange rate given an increase in imports relative to exports. Under
fixed exchange rates, the monetary authority intervenes to prevent this appreciation
by expanding the money supply. Figure 7 shows the response of such variables to a
shock in government spending. In the case of the flexible exchange rate countries, we
observe an appreciation of the real exchange rate as the theory predicts. In the case of
the countries operating under fixed exchange rate regimes we do not observe monetary
accommodation. We observe an initial real exchange rate depreciation that disappears
after the first quarter that may be explained by the mechanism proposed by Ravn et al.
(2012) as explained before.
We next analyze the response of economies that differ in their degree of trade open-
ness. Figure 8 shows impact multipliers that are statistically positive and different
12
between both cases (0.62 for closed economies and 0.37 for open economies). Long
run multipliers have positive point estimates, but they are statistically positive only for
closed economies. Although they are statistically not distinguishable from each other
after the first year, the uncertainty is higher in the case of the open economies. Those
results are consistent with IMV, although our point estimates are higher and nonnega-
tive in both cases while they find negative multipliers in the case of the open economies.
Those results are also consistent in principle with the Mundell Fleming model. In the
case of open economies, this model predicts that part of the demand generated by an
increase in government spending should go to imported goods, ameliorating the domestic
output response. This also implies an appreciation of the real exchange rate, which is
what we observe in figure 9 after the third quarter.
Figure 10 shows the differences between the fiscal multipliers of high-income vs. de-
veloping countries. We find that the impact multipliers are both positive and not sta-
tistically different. In the long run, the fiscal multiplier for developing countries is 0.88,
suggesting some degree of crowding out because output rises less than total government
spending. This multiplier is statistically different from zero. In the case of high-income
countries, we find that the fiscal multiplier is positive although not statistically different
from zero. Those results are in contrast to IMV, who find a negative multiplier not
statistically different from zero for developing countries and positive multipliers for high
income countries. Interestingly, figure 11 shows that the fiscal multipliers are both sta-
tistically different form zero during the first 3 years if we do not control for either the
interest rate or the real exchange rate, suggesting the importance of those variables. The
same figure shows that if we do not control for the interest rate or the real exchange rate,
impact multipliers are both positive and statistically different. To precisely analyze the
mechanisms behind this behavior, figure 12 shows how in the case of developing coun-
tries the interest rate increases after an initial negative response, suggesting an initial
accommodation from the monetary authority. In the case of the high-income countries,
there is also an initial accommodation of the monetary authority, but it is zero after
the third quarter throughout all the period of analysis. In the case of the real exchange
rate, the only change that is distinguishable from zero is in the case of the developing
13
countries, where we can observe a depreciation of this index in the first quarters. The
picture that emerges is that the crowding out effect, although present in both cases, is
bigger in the case of high income countries, a result in contrast with previous studies.
We think those results are more in line with the notion that developing countries have
more binding constraints to spending that can be alleviated with fiscal stimulus.
The level of indebtedness may influence the effect of government spending on output.
The intuition is that a high level of debt may affect the expectations about repayment
and about future fiscal adjustment, counteracting the expansionary effects of an increase
in government spending. Figure 13 shows that the fiscal stimulus is more effective in
countries with a low level of debt: the multiplier in the long run is 1.49 in the case of
low debt countries vs. 0.39 in the case of high debt countries, and the impact multiplier
is 0.44 vs 0.37, respectively. After the third quarter, however, the multiplier for high
debt countries is not different from zero, while it is zero in in the case of low debt
countries. Although changes in the interest rate and the real exchange rate are not
statistically different from zero (figure 14), the point estimates suggest that an increase
in government spending in a high debt country may either signal difficulty of payment
later or inflationary concerns, as the increase in the interest rate suggests.
5.1 Sources of divergence with previous results
We investigate in this section the factors that explain the divergence with previous
results, specifically the differences with the results of IMV. In order to make the com-
parisons possible through all the empirical exercises of this section, we do not use per
capita data as we have done so far in the paper and instead we use aggregate data, we
calculate the multiplier discounting it by the median interest rate as IMV, and we use the
same variables that they use in the estimation, changing the discount rate that we used
in previous estimations with the current account measure. We take the IMV dataset
from the public version posted with their publication. Using the IMV dataset, we are
able to replicate the results for each of the cases reported in their paper. We calculate
the aggregate multipliers in their case to facilitate comparissons with our results.
14
We start by calculating the average multipliers for all the IMV sample of countries
using an OLS estimator (this is the original estimation in their paper) and the panel
SVAR estimator that we use in this paper. Figure 15 shows that the biases in the point
estimates are considerable in the long run: while the OLS estimator produces an average
multiplier 0.26, the panel SVAR estimator produces a multiplier of -0.4. The bias reflects
that the error term capturing shocks to output is positively correlated with government
spending, as expected. The confidence intervals overlap, and the GMM estimator has
much bigger standard errors due to the bootstrapping procedure we use. The fact that
the panel SVAR estimator produces even lower multipliers than the OLS estimator does
not explain the differences in the results, given that although we use this estimator, we
generally obtain much higher multipliers than Ilzetzki et al. If anything, the use of this
estimator reduces the differences between our results and their results.
We next analyze the effect of having different countries in the samples. As we will
show, this is the key factor explaining the differences in results. We create a panel with
the IMV data that only considers countries present in our data. In total, 34 countries are
present in both datasets. The countries present in the IMV dataset but not in ours are are
Botswana, Bulgaria, Croatia, Estonia, Israel, Latvia, Lithuania, Romania, Slovenia and
Uruguay. The countries present in our data set but not in the IMV dataset are Austria,
Bolivia, China, Costa Rica, Guatemala, Hong Kong, India, Indonesia, Japan, Korea,
Luxembourg, New Zealand, Pakistan, Paraguay, Philippines, the Russian Federation,
Singapore, Switzerland, Taiwan, Venezuela and Vietnam. Figure 16 shows the results
of estimating the fiscal multipliers with this common set of countries and compares it
with the multipliers estimated with all the countries present in each dataset. The upper
panel compares the fiscal multipliers when we use the original set of countries and the
panel SVAR estimator. As we have shown before, the IMV sample produces a multiplier
of 0.05 in the short run and -0.40 in the long run (not statistically different from zero),
while we obtain a multiplier of 0.43 in the short run and of 0.83 in the long run (both
statistically different from zero).18 The lower panel shows the same estimation with the
18Those numbers differ slightly from the previously reported multipliers because we use the currentaccount instead of the discount rate and because we use aggregate data instead of per capital data.
15
common set of countries. Results are very similar: Using the IMV data, we obtain a
fiscal multiplier of 0.40 and 0.43 in the short and in the long run, statistically different
from zero through almost in the entire period. We obtain with our data fiscal multipliers
of 0.52 and 0.54 in the short and the long run. The confidence intervals vastly overlap.
It is clear from this figure that the sample composition explains almost all the differences
between our results and the results of IMV.
When we use the same sample of countries present in both datasets to analyze the
multipliers in developing vs high income countries, we still obtain positive fiscal mul-
tipliers in the short run, a result that is different to their results, although it is lower
than the multipliers of high income countries and zero in the long run as it is shown in
Figure 17. In the other cases (debt, exchange rate regime and degree of trade openness),
we obtained similar results than the ones at the beginning of section 5. 19. Table 6
summarizes the results presented in figures 15 and 16, namely that sample selection is
the main driver of the differences between our estimates and IMV results.
6 Conclusions
We use quarterly data in a panel of 55 countries between 1988 and 2010 to analyze the
effect of shocks to government spending on GDP and private consumption. In addition
to using a more comprehensive dataset than previous studies, as an additional innovation
we use panel SVAR techniques to correct for the correlation of the explanatory variables
with the error terms known to be present in this type of settings. This helps to ameliorate
other possible sources of endogeneity like anticipation effects of third variables that may
affect both output and government spending.
We find positive multipliers of around 0.3 on impact and between 0.9 and 1.0 in the
long run. In addition, we find positive private consumption and employment responses,
giving support to Keynessian theories. Those numbers, although not larger than one,
19We also changed the time period and dropped the crisis years between 2008 and 2010. The periodof analysis did not have almost any effect on the results.
16
show that fiscal policy can be effective stimulating the economy, an important policy
issue especially in times of economic turmoil. A short run-multiplier bigger than zero
gives support to a countercyclical fiscal policy. A long-run multiplier bigger than zero
is important to ensure that countercyclical fiscal policy does not have negative long-
run effects. At the end, with the possibility that fiscal multipliers are lower than one
(although positive) policymakers should weight the benefits of fiscal policy under times
of economic stress versus the potential displacement of private investment that could
have other long-run effects.
Results in this paper question previous conclusions drawn in the panel VAR literature
that analyzes fiscal multipliers. The type of estimator and the sample selection are
important drivers of those conclusions. Contrary to previous research,we find that the
fiscal multiplier is larger in developing than in high-income countries, that it is zero in
high debt countries and in countries operating under flexible exchange rates (instead
of negative), and we do not find strong evidence of monetary accommodation. The
differences between our results and those of previous studies come fundamentally from
the sample of countries. The use of a GMM estimator instead of an OLS estimator also
shows that OLS estimators induce some important biases in the results.
17
References
[1] Alesina, Alberto and Ardagna, Silvia, 1998. “Tales of Fiscal Adjustment,” Economic
Policy, vol. 0(27), pages 73 - 86.
[2] Auerbach, Alan and Yuriy Gorodnichenko, 2012. “Measuring the Output Responses
to Fiscal Policy,” American Economic Journal: Economic Policy, American Economic
Association, vol. 4(2), pages 1 - 27.
[3] Barro, Robert and Charles Redlick, 2011. “Macroeconomic Effects from Government
Purchases and Taxes,” Quarterly Journal of Economics, vol. 127(3), pages 829 - 887.
[4] Barry, Frank and Devereux, Michael, 2013. “Expansionary Fiscal Contraction: A
Theorethical Exploration,” Journal of Macroeconomics, vol. 25(1), pages 1 - 23.
[5] Beetsma, Roel and Massimo Giuliodori, 2011. “The Effects of Government Purchases
Shocks: Review and Estimates for the EU,” Economic Journal, Royal Economic So-
ciety, vol. 121(550), pages F4 - F32.
[6] Blanchard, Olivier and Roberto Perotti, 2002. “An Empirical Characterization of
the Dynamic Effects of Changes in Government Spending and Taxes on Output,”
Quarterly Journal of Economics (November), pages 1329 - 1368.
[7] Edelberg, Wendy, Martin Eichenbaum and Jonas D. M. Fisher, 1999. “Understanding
the Effects of a Shock to Government Purchases,” Review of Economic Dynamics 2
(January), pages 166 - 206.
[8] Fatıs, Antonio and Mihov, Ilian, 2001. “The Effects of Fiscal Policy on Consumption
and Employment: Theory and Evidence,” CEPR Discussion Papers 2760.
[9] Giavazzi, Francesco and Pagano, Marco, 1990. “Can Severe Fiscal Contractions be
Expansionary? Tales from two Small European Economies,” NBER Macroeconomics
Annual.
18
[10] Holtz-Eakin, Douglas and Newey, Whitney and Rosen, Harvey S, 1988. “Estimating
Vector Autoregressions with Panel Data,” Econometrica, Econometric Society, vol.
56(6), pages 1371 - 95.
[11] Ilzetzki, Ethan, Carlos Vegh and Enrique Mendoza, 2013 “How Big (Small?) are
Fiscal Multipliers?,” Journal of Monetary Economics, vol. 60(2), pages 239 - 254.
[12] Ilzetzki, Ethan, Carmen Reinhart and Kenneth Rogoff, 2008. “De-facto exchange
rate classification 1945-2010” in “Exchange Rate Arrangements Entering the 21st
Century: Which Anchor Will Hold?”
[13] Judson, Ruth A. and Ann L. Owen, 1999 “Estimating Dynamic Panel Data Models:
A Guide for Macroeconomists,” Economic Letters, vol. 65, pages 9 - 15.
[14] Love, Inessa and Zicchino, Lea, 2006. “Financial development and dynamic invest-
ment behavior: Evidence from panel VAR,” The Quarterly Review of Economics and
Finance, Elsevier, vol. 46(2), pages 190 - 210.
[15] Merten, Karel and Morten Ravn, 2012. “Empirical evidence on the aggregate effects
of anticipated and unanticipated U.S. tax policy shocks,” American Economic Journal:
Economic POlicy, vol 4(2), pages 27 - 64
[16] Mountford, Andrew and Harald Uhlig, 2009. “What are the effects of fiscal policy
shocks?,” Journal of Applied Econometrics, vol. 24(6), pages 960 - 992.
[17] Nakamura, Emi and Jon Steinsson, 2011. “Fiscal Stimulus in a Monetary Union,”
NBER Working paper 17391.
[18] Perotti, Roberto, 2005. “Estimating the Effects of Fiscal Policy in OECD Coun-
tries,” CEPR Discussion Paper 4842.
[19] Ramey, Valerie A. and Matthew Shapiro, 1998. “Costly Capital Reallocation and
the Effects of Government Spending,” Carnegie Rochester Conference on Public Pol-
icy.
19
[20] Ramey, Valerie A., 2011. “Identifying government spending shocks: it’s all in the
timing,” Quarterly Journal of Economics, vol. 126(1), pages 1 - 50.
[21] Ravn, Morten, Stephanie Schmitt-Grohe and Martın Uribe, 2012. “Explaining the
Effects of Government Spending Shocks,” Journal of Monetary Economics, vol. 59(3),
pages 215 - 34.
[22] Reinhart, Carmen and Kenneth Rogoff, 2010. Tables in “From Financial Crisis to
Debt Crises,” NBER WP 15795.
[23] Suarez-Serrato, Juan Carlos and Philippe Wingender, 2011. “Estimating Local Fis-
cal Multipliers,” Working paper, University of California, Berkeley.
20
Table 1: Countries included in the sample 1988q1 to 2010q4
1 ARGENTINA 29 KOREA2 AUSTRALIA 30 LUXEMBOURG3 AUSTRIA 31 MALAYSIA4 BELGIUM 32 MEXICO5 BOLIVIA 33 NETHERLANDS6 BRAZIL 34 NEW ZEALAND7 CANADA 35 NORWAY8 CHILE 36 PAKISTAN9 CHINA,P.R.: MAINLAND 37 PARAGUAY
10 COLOMBIA 38 PERU11 COSTA RICA 39 PHILIPPINES12 CZECH REPUBLIC 40 POLAND13 DENMARK 41 PORTUGAL14 ECUADOR 42 RUSSIA15 EL SALVADOR 43 SINGAPORE16 FINLAND 44 SLOVAK REPUBLIC17 FRANCE 45 SOUTH AFRICA18 GERMANY 46 SPAIN19 GREECE 47 SWEDEN20 GUATEMALA 48 SWITZERLAND21 CHINA,P.R.:HONG KONG 49 TAIWAN22 HUNGARY 50 THAILAND23 ICELAND 51 TURKEY24 INDIA 52 UNITED KINGDOM25 INDONESIA 53 UNITED STATES26 IRELAND 54 VENEZUELA, REP. BOL.27 ITALY 55 VIETNAM28 JAPAN
21
Table 2: Basic Statistics
GDP Private Government Government spending /consumption spending GDP
# obs 3890 3904 3905 3890All Mean 198679 122002 46654 0.161
Std. Dev. 443017 294976 97318 0.054# obs 1666 1680 1681 1666
Fixed Mean 319119 200646 72223 0.162Std. Dev. 635755 426306 136690 0.050
# obs 2224 2224 2224 2224Flexible Mean 108458 62594 27328 0.161
Std. Dev. 146995 85437 40550 0.057# obs 1950 1950 1950 1950
Open Mean 78728 43810 18174 0.166Std. Dev. 112579 66055 25729 0.059
# obs 1769 1767 1768 1769Closed Mean 347385 217367 81254 0.161
Std. Dev. 613942 412542 133545 0.048# obs 2611 2625 2627 2611
low debt Mean 156294 93879 38685 0.162Std. Dev. 291623 187439 72623 0.053
# obs 937 937 936 937high debt Mean 374347 236476 84121 0.172
Std. Dev. 730064 495247 150066 0.050# obs 2213 2213 2212 2213
high income Mean 298379 182429 72076 0.191Std. Dev. 563616 378019 121999 0.043
# obs 1677 1691 1693 1677developing Mean 67113 42921 13439 0.122
Std. Dev. 75577 53606 21316 0.040
All quantities in millions of 2008 dollars
22
Table 3: Episodes of flexible and fixed exchange ratesCountry Periods of Fixed exchange rate. All other periods are considered flexible regimes
1 Argentina 1993q1:2001q3;2007q1-2010q42 Australia -3 Austria 1988q1 - 2010q44 Belgium 1988q1 - 2010q45 Bolivia 1988q1 - 2010q46 Brazil 1989q1; 1994q3-1998q47 Canada 1988q1-2001q48 Chile 1989q2 - 1991q4; 1998q3 - 1999q29 China 1992q3 - 2005q2; 2008q4 - 2010q410 Colombia -11 Costa rica 1991q1 - 2010q412 Czech Republic 1988q1 - 1995q4; 1997q2 -2001q413 Denmark 1988q1 - 2010q414 Ecuador 1997q1 - 1997q3;2000q1 - 2010q415 El Salvador 1990q2 - 2010q416 Finland 1988q1 - 1992q2, 1993q2 - 2010q417 France 1988q1 - 2010q418 Germany 1999q1 - 2010q419 Greece 1988q1 - 2010q420 Guatemala 1988q3 - 1989q1;1991q2 - 2010q421 Hong Kong 1988q1 - 2010q422 Hungary 1994q2 - 2005q1; 2009q423 Iceland 1988q1:2000q324 India 1988q1 - 2010q425 Indonesia 1988q1 - 1997q226 Ireland 1988q1 - 1997q227 Italy 1988q1 - 1992q2; 1993q2 - 2010q428 Japan -29 South Korea 1988q1 - 1997q330 Luxembourg 1988q1 - 2010q431 Malaysia 1988q1 - 1997q2; 1998q4 - 2007q432 Mexico 1988q4 - 1994q433 Netherlands 1988q1 - 2010q434 New zealand -35 Norway -36 Pakistan 1988q1 - 2007q4; 2008q3 - 2010q437 Paraguay 1991q1 - 1999q2; 2010q1 - 2010q438 Peru 1993q4 - 2010q439 Philippines 1988q1 - 1993q1; 1995q3 - 1997q2; 1999q4 - 2007q340 Poland 1990q1 - 1991q141 Portugal 1988q1 - 2010q442 Russia 1988q1 - 1991q4; 1999q4 - 2009q343 Singapore -44 Slovakia 1988q1 - 1992q4; 1993q2 - 1997q2; 1998q4 - 2010q445 South africa -46 Spain 1988q1 - 2010q447 Sweden 1988q1 - 1992q348 Switzerland 1988q1 - 1998q449 taiwan 1988q1 - 2010q450 thailand 1988q1 - 1997q251 Turkey -52 United Kingdom 1990q4 -1992q253 United States -54 Venezuela 1996q3 - 2007q3; 2008q1 - 2010q455 Vietnam 1988q1 - 2010q4
Source: Own calculations based on Ilzetzki, Reinhart and Rogoff tables (2008)
23
Table 4: Open and Closed Economies∗Country Periods when open. All other periods are classified as closed regimes
1 Argentina No data before 1993q1. Closed all other periods2 Australia -3 Austria 1988q1 -2010q44 Belgium No data before 1995q1. Open 1988q1 - 1994q4; 2008q1 -2008q35 Bolivia 2004q4 - 2010q46 Brazil -7 Canada 1993q3 - 2009q1 and 2010q28 Chile 1990q1 - 1991q1;1995q4; 2000q3 - 2010q49 China No data before 1992q2; open 2004q4 - 2008q110 Colombia -11 Costa rica no data before 1991q1. open 1991q1 - 2010q412 Czech Republic no data before 1994q4; open 1995q1 - 2010q413 Denmark no data before 1989q4; open 1990q1 - 2010q414 Ecuador 1991q4 - 2010q415 El Salvador no data before 1990q4; open 2000q2 - 2000q4;2003q3; 2004q2 - 2004q4;
2005q2; 2006q1 - 2006q3; 2007q1 - 2008q3; 2010q1 - 2010q416 Finland 2004q1 - 2010q417 France -18 Germany 2000q1 - 2010q419 Greece 2000q1 -2000q2; 20008q220 Guatemala 2001q1 - 2008q3; 2010q2 - 2010q421 Hong Kong 1988q1 -2010q422 Hungary no data before 1995q2; open 1995q3 - 2010q423 Iceland no data before 1996q4; open 1997q1 - 2010q424 India -25 Indonesia no data before 1990q1; open 1997q4 - 1998q4; 2000q1 - 2002q4;2008q1-2008q226 Ireland no data before 1996q4; open 1997q1 - 2010q427 Italy -28 Japan -29 South Korea 1988q1 - 1988q4; 1997q1 - 2010q230 Luxembourg -31 Malaysia No data32 Mexico -33 Netherlands 1988q1 -2010q434 New Zealand 1992q4; 1999q4 - 2002q4; 2004q2; 2006q2 -2006q4; 2008q1 - 2009q135 Norway 1988q1 -2010q436 Pakistan -37 Paraguay no data before 1993q4; open 1994q1 - 2010q438 Peru -39 Philippines 1990q4 - 2010q240 Poland 2002q3 - 2010q441 Portugal 1988q1 - 2010q442 Russia 2003q143 Singapore No data44 Slovakia no data before 1994q4; open 1995q1 - 2010q445 South Africa 1988q1 - 2010q446 Spain 2000q2 - 2001q2; 2006q4 - 2008q247 Sweden No data before 1993q1; open 1993q2 - 2010q448 Switzerland 1988q1 - 2010q449 Taiwan 1988q1 - 2010q450 Thailand No data before 1992q4; open 1993q1 - 2010q451 Turkey -52 United Kingdom 2008q2 - 2008q3; 2010q353 United States -54 Venezuela No data before 1997q4: open 2006q2 - 2009q155 Vietnam No data before 1989q4: open 1990q1 - 1990q4; 1994q1 - 2010q4
∗ Open economies are considered those with a Trade (nominal exports+imports) to GDP ratiogreater than 60%Source: Own calculations
24
Table 5: Episodes of High Debt1
Country Episodes of high debt
1 Argentina 1988q1 - 1989q4; 2002q1 - 2006q42 Australia -3 Austria 2009q1 - 2010q44 Belgium 1988q1 - 2010q45 Bolivia 1988q1 - 1995q4;1998q1 - 2005q46 Brazil 1988q1 - 1988q4; 1991q1 - 1993q4; 1997q1 - 1998q4;
2001q1 - 2001q4; 2003q1 - 2004q4; 2010q47 Canada 1988q1 - 2002q48 Chile 1988q1 - 1988q49 China -10 Colombia -11 Costa rica 1988q1 - 1991q4; 2003q1 - 2004q412 Czech Republic No data13 Denmark 1988q1 - 1999q414 Ecuador 1988q1 - 1996q4; 1988q1 - 2001q415 El Salvador -16 Finland 1995q1 - 1997q417 France 2003q1 - 2010q418 Germany -19 Greece 1988q1 - 2010q420 Guatemala -21 Hong Kong No data22 Hungary No data before 1990q4; 1991q1 - 2003q4; 2006q1 - 2010q423 Iceland 2009q1 - 2010q424 India -25 Indonesia 1998q1 -2002q426 Ireland 1988q1 - 1996q4; 1988q1 - 2001q427 Italy 1988q1 - 2010q428 Japan 1995q1 - 2010q429 South Korea -30 Luxembourg No data31 Malaysia 1988q1 - 1992q432 Mexico 1988q1 - 1989q433 Netherlands 1992q1 - 1993q434 New Zealand No data before 1991q4; 1992q1 - 1993q435 Norway -36 Pakistan No data37 Paraguay -38 Peru 1990q1 - 1994q439 Philippines 1988q1 - 1995q4; 1997q1 - 2006q440 Poland No data before 1989q4; 1990q1 -1994q441 Portugal 2003q1 - 2010q442 Russia No data before 1989q4; 1990q1 -1993q4; 1999q1 - 1999q443 Singapore 1988q1 - 2009q444 Slovakia No data before 1992q445 South Africa -46 Spain 1993q1 - 2000q447 Sweden -48 Switzerland -49 Taiwan -50 Thailand -51 Turkey 2002q1-2005q452 United Kingdom 2010q1 - 2010q453 United States 1991q1 - 1999q4; 2003q1 - 2010q454 Venezuela 1989q1 - 1995q455 Vietnam No data
Gross Debt of Central Government Exceeding 60% of GDPSource: Own calculations based on Reinhart and Rogoff (2010)
25
Table 6: Fiscal Multipliers Calculated Using Different Estimators and DatasetsMultipliers
Dataset Estimator Sample Impact Long-RunOriginal IMV IMV OLS All Countries in IMV -0.02 0.26
IMV Sample Selection IMV OLS footnote (1) 0.3 ∗ 0.7∗
IMV GMM IMV GMM All countries in IMV 0.02 -0.40IMV GMM + Sample selection IMV GMM footnote (1) 0.40∗ 0.43 ∗
Original CB CB GMM All countries in CB 0.43∗ 0.83∗
CB Sample selection CB GMM footnote (2) 0.52 ∗ 0.54+
(1) Countries excluded from the IMV dataset are Botswana, Bulgaria, Croatia, Estonia, Israel,Latvia, Lithuania, Romania, Slovenia and Uruguay. The final dataset is comprised of the countriesthat intersect with the CB dataset(2) Countries excluded from the CB dataset are Austria, Bolivia, China, Costa Rica, Guatemala,Hong Kong, India, Indonesia, Japan, Korea, Luxembourg, New Zealand, Pakistan, Paraguay, Philip-pines, Russian Federation, Singapore, Switzerland, Taiwan, Venezuela, Vietnam. The final datasetis comprised the countries that intersect with the IMV dataset(3) ∗ Statistically different from zero at 5% of significance. + Statistically different from zero at 10%of significance.
26
Figure 1: Aggregate Fiscal Multipliers
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
1 3 5 7 9 11 13 15 17 19
Impact: 0.36
Long-run: 0.92
The SVAR system includes log of per-capita government spending, log of per-capita output, policy(discount) interest rate and an index of the real exchange rate. Total number of observations was 3131for all 55 countries
27
Figure 2: Impulse response functions from a shock to government spending
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
1 3 5 7 9 11 13 15 17 19 21
Government Spending
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1 3 5 7 9 11 13 15 17 19 21
Output
-20.0
-10.0
0.0
10.0
20.0
30.0
40.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1 3 5 7 9 11 13 15 17 19 21
Real exchange rate
1. The impulse response function of the variable X with respect to the variable Y is the response ofthe variable X to a one unit standard deviation shock in the variable Y. The units are in percentagepoints. The confidence intervals represent 90% of the distribution (5% the lower and 95% the higher)2. 55 countries included, 3131 observations
28
Figure 3: Aggregate Fiscal Multipliers Adding Consumption
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
1 3 5 7 9 11 13 15 17 19
Impact: 0.38
Long-run: 1.00
The SVAR system includes log of per-capita government spending, log of per-capita output, policy(discount) interest rate, an index of the real exchange rate and log of per capita private consumption.Total number of observations was 3065 for all 55 countries
29
Figure 4: Impulse response functions adding consumption
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
1 3 5 7 9 11 13 15 17 19 21
Government Spending (shock)
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1 3 5 7 9 11 13 15 17 19 21
Output
-15.0
-10.0
-5.0
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1 3 5 7 9 11 13 15 17 19 21
Real Exchange Rate
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1 3 5 7 9 11 13 15 17 19 21
Consumption
1. The impulse response function of the variable X with respect to the variable Y is the response ofthe variable X to a one unit standard deviation shock in the variable Y. The units are in percentagepoints. The confidence intervals represent 90% of the distribution (5% the lower and 95% the higher)
2. 55 countries included, 3065 observations
30
Figure 5: Fiscal Multipliers: Fixed vs. Flexible Exchange Rate Regimes
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
1 3 5 7 9 11 13 15 17 19
Flexible
Fixed
Impact: 0.57
Impact: 0.27
Long-run: 1.58
Long-run: -0.27
The SVAR system includes log of per-capita government spending and log of per-capita output. Totalnumber of observations was 2018 for fixed exchange rate regimes and 1482 for flexible exchange rateregimes
31
Figure 6: Fiscal Multipliers: Flexible vs. Fixed exchange rate regimes Controlling forReal Interest Rate and Real Exchange Rate
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
1 3 5 7 9 11 13 15 17 19
Flexible
Fixed
Impact: 0.41
Impact: 0.29
Long-run: 0.54
Long-run: 1.06
Discount rate and real exchange rate included
The SVAR system includes log of per-capita government spending, log of per-capita output, policy(discount) interest rate and an index of the real exchange rate. Total number of observations was 1516for fixed exchange rate regimes and 1469 for flexible exchange rate regimes
32
Figure 7: Impulse response functions: Flexible and fixed exchange rate regimes
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
1 3 5 7 9 11 13 15 17 19 21
Government spending
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
1 3 5 7 9 11 13 15 17 19 21
Government Spending
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1 3 5 7 9 11 13 15 17 19 21
Output
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1 3 5 7 9 11 13 15 17 19 21
Output
-30.0
-20.0
-10.0
0.0
10.0
20.0
30.0
40.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-20.0
-10.0
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1 3 5 7 9 11 13 15 17 19 21
Real exchange rate
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1 3 5 7 9 11 13 15 17 19 21
Real exchange rate
1. The impulse response function of the variable X with respect to the variable Y is the response ofthe variable X to a one unit standard deviation shock in the variable Y. The units are in percentagepoints. The confidence intervals represent 90% of the distribution (5% the lower and 95% the higher)2. The responses of Fixed exchange rate regimes are on the right
33
Figure 8: Fiscal Multipliers: Open vs. Closed economies
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1 3 5 7 9 11 13 15 17 19
Closed
Open
Impact: 0.62
Impact: 0.27
Long-run: 0.23
Long-run: 1.76
Discount rate and real exchange rate included
The SVAR system includes log of per-capita government spending and log of per-capita output, policy(discount) interest rate and an index of the real exchange rate. Total number of observations was 1542for open economies and 1370 for closed economies.
34
Figure 9: Impulse response functions: Open vs. closed economies
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
1 3 5 7 9 11 13 15 17 19 21
Government spending
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1 3 5 7 9 11 13 15 17 19 21
Government spending
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1 3 5 7 9 11 13 15 17 19 21
Output
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1 3 5 7 9 11 13 15 17 19 21
Output
-40.0
-30.0
-20.0
-10.0
0.0
10.0
20.0
30.0
40.0
50.0
60.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-40.0
-30.0
-20.0
-10.0
0.0
10.0
20.0
30.0
40.0
50.0
60.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
1 3 5 7 9 11 13 15 17 19 21
Real exchange rate
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
1 3 5 7 9 11 13 15 17 19 21
Real exchange rate
1. The impulse response function of the variable X with respect to the variable Y is the response ofthe variable X to a one unit standard deviation shock in the variable Y. The units are in percentagepoints. The confidence intervals represent 90% of the distribution (5% the lower and 95% the higher)2. Closed economies responses on the right
35
Figure 10: Fiscal Multipliers: High Income vs. Developing countries
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
1 3 5 7 9 11 13 15 17 19
High Income
Developing
Impact: 0.39
Impact: 0.36
Long-run: 0.88
Long-run: 0.38
Discount rate and real exchange rate included
The SVAR system includes log of per-capita government spending and log of per-capita output, policy(discount) interest rate and an index of the real exchange rate. Total number of observations was 2026for high income countries and 1640 for developing countries.
36
Figure 11: Fiscal Multipliers: High Income vs. Developing countries
-0.5
0.0
0.5
1.0
1.5
2.0
1 3 5 7 9 11 13 15 17 19
High Income
Developing
Impact: 0.34
Impact: 0.60
Long-run: 0.84
Long-run: 0.64
The SVAR system includes log of per-capita government spending and log of per-capita output.Different from Figure 10, this figure shows a VAR that does not include interest rate nor the realexchange rate. Total number of observations was 1831 for high income countries and 1300 fordeveloping countries.
37
Figure 12: Impulse response functions: High Income vs Developing countries
0.0
0.5
1.0
1.5
2.0
2.5
1 3 5 7 9 11 13 15 17 19 21
Government spending
0.0
1.0
2.0
3.0
4.0
5.0
6.0
1 3 5 7 9 11 13 15 17 19 21
Government spending
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
1 3 5 7 9 11 13 15 17 19 21
Output
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1 3 5 7 9 11 13 15 17 19 21
Output
-30.0
-25.0
-20.0
-15.0
-10.0
-5.0
0.0
5.0
10.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-40.0
-20.0
0.0
20.0
40.0
60.0
80.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1 3 5 7 9 11 13 15 17 19 21
Real exchange rate
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1 3 5 7 9 11 13 15 17 19 21
Real exchange rate
1. The impulse response function of the variable X with respect to the variable Y is the response ofthe variable X to a one unit standard deviation shock in the variable Y. The units are in percentagepoints. The confidence intervals represent 90% of the distribution (5% the lower and 95% the higher)2. Developing countries’ responses on the right
38
Figure 13: Fiscal Multipliers: High Debt vs. Low Debt countries controlling for the realinterest rate and an index of the real exchange rate
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1 3 5 7 9 11 13 15 17 19
High Debt
Low Debt
Impact: 0.44
Impact: 0.37
Long-run: 0.39
Long-run: 1.49
Discount rate and real exchange rate included
The SVAR system includes log of per-capita government spending and log of per-capita output, policy(discount) interest rate and an index of the real exchange rate. Total number of observations was 2082for low debt and 712 for high debt countries.
39
Figure 14: Impulse response functions: High debt vs low debt countries
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
1 3 5 7 9 11 13 15 17 19 21
Government spending
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
1 3 5 7 9 11 13 15 17 19 21
Government spending
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
1 3 5 7 9 11 13 15 17 19 21
Output
-0.5
0.0
0.5
1.0
1.5
2.0
1 3 5 7 9 11 13 15 17 19 21
Output
-40.0
-30.0
-20.0
-10.0
0.0
10.0
20.0
30.0
40.0
50.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-70.0
-60.0
-50.0
-40.0
-30.0
-20.0
-10.0
0.0
10.0
20.0
30.0
1 3 5 7 9 11 13 15 17 19 21
Discount rate
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1 3 5 7 9 11 13 15 17 19 21
Real exchange rate
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
1 3 5 7 9 11 13 15 17 19 21
Real exchange rate
1. The impulse response function of the variable X with respect to the variable Y is the response ofthe variable X to a one unit standard deviation shock in the variable Y. The units are in percentagepoints. The confidence intervals represent 90% of the distribution (5% the lower and 95% the higher)2. Low debt countries responses on the right
40
Figure 15: Method comparison: OLS vs. Dynamic Panel GMM
-1.2
-0.8
-0.4
0.0
0.4
0.8
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Impact: -0.016 Long Run: 0.26
All countries in IMV sample, OLS estimator
-1.2
-0.8
-0.4
0.0
0.4
0.8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Impact: 0.02
Long-run: -0.40
All countries in IMV sample, Dynamic panel GMM
The upper figure shows the multipliers that we obtained when we used all countries in the Ilzetzki,Mendoza and Vegh (2013) dataset and an OLS estimator. The lower figure shows the multipliers thatwe obtained using the same countries and the dynamic panel GMM estimator proposed by Holtz-Eakinet al. (1988)
41
Figure 16: Sample selection effect
-1.3
-0.8
-0.3
0.2
0.7
1.2
1 3 5 7 9 11 13 15 17 19
Impact: 0.02
Long-run: -0.40
IMV sample, 44 countries
-1.3
-0.8
-0.3
0.2
0.7
1.2
1 3 5 7 9 11 13 15 17 19
Impact: 0.43
Long-run: 0.83
CB 55 countries
-1.3
-0.8
-0.3
0.2
0.7
1.2
1 3 5 7 9 11 13 15 17 19
Impact: 0.40
Long-run: 0.43
IMV Sample, 34 countries (intersection with CB)
-1.3
-0.8
-0.3
0.2
0.7
1.2
1 3 5 7 9 11 13 15 17 19
Impact: 0.52 Long-run: 0.54
CB Sample, 34 countries (intersection with IMV)
All estimations use the dynamic panel GMM estimator proposed by Holtz-Eakin et al. (1988) Theupper left figure shows the multipliers calculated using the complete Ilzetzki et al. dataset (IMV). Theupper right figure shows the multipliers calculated using all countries in our sample (CB). The lowerleft figure shows the multipliers calculated with the IMV dataset but using only the countries present inboth the IMV dataset and in our dataset (CB). The lower right figure shows the multipliers calculatedwith our dataset but using only the countries present in both the IMV dataset and in our dataset(CB). The countries taken out of the original IMV dataset were Botswana, Bulgaria, Croatia, Estonia,Israel, Latvia, Lithuania, Romania, Slovenia and Uruguay. The countries taken out of our originaldataset were Austria, Bolivia, China, Costa Rica, Guatemala, Hong Kong, India, Indonesia, Japan,Korea, Luxembourg, New Zealand, Pakistan, Paraguay, Philippines, Russian Federation, Singapore,Switzerland, Taiwan, Venezuela, Vietnam.
42
Figure 17: Fiscal Multipliers: High Income vs. Developing countries using IMV sample
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1 3 5 7 9 11 13 15 17 19
High Income
Developing
Impact: 0.5
Impact: 0.2
Long-run: 0.6
Long-run: --1.3
The estimation uses GMM and a sample of countries from the IMV dataset that intersects with ourdataset. Total number of observations was 1290 for high income countries and 613 for developingcountries.
43
SU
PPLEM
EN
TA
RY
INFO
RM
AT
ION
:D
ATA
SO
UR
CES
AN
DT
IME
PER
IOD
S
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
Arg
enti
na
GD
Pat
curr
ent
pri
ces
Secre
tarı
ade
Pro
gra
macio
nEconom
ica
Min
iste
rio
de
Econom
ıay
Obra
sy
Serv
icio
sPublicos
1993q1
-2010q2
Public
consu
mpti
on
Secre
tarı
ade
Pro
gra
macio
nEconım
ica
Min
iste
rio
de
Econom
iay
Obra
sy
Serv
icio
sPublicos
1993q1
-2010q2
Pri
vate
consu
mpti
on
Secre
tarı
ade
Pro
gra
macio
nEconım
ica
Min
iste
rio
de
Econom
ıay
Obra
sy
Serv
icio
sPublicos
1993q1
-2010q2
GD
Pim
plicit
deflato
r(1
993=
100)
Inst
ituto
Nacio
nalde
Est
adıs
tica
yC
enso
s.1993q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1991q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1994q1
-2010q4
Em
plo
ym
ent
Inst
ituto
Nacio
nalde
Est
adıs
tica
yC
enso
s.1991q1
-2010q1
Aust
ralia
GD
Pat
curr
ent
pri
ces
Aust
ralian
Bure
au
ofSta
tist
ics
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Public
Aust
ralian
Bure
au
ofSta
tist
ics
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Pri
vate
Aust
ralian
Bure
au
ofSta
tist
ics
1988q1
-2010q2
GD
Pdeflato
r(Q
3/2005-Q
2/2006=
100)
Aust
ralian
Bure
au
ofSta
tist
ics
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
(Thousa
nds)
Aust
ralian
Bure
au
ofSta
tist
ics
1988q1
-2010q3
Aust
ria
GD
Pat
curr
ent
pri
ces
Sta
tist
ikA
ust
ria
1988q1
-2010q2
Govern
ment
spendin
gIS
FD
ATA
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
Sta
tist
ikA
ust
ria
1988q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Sta
tist
ikA
ust
ria
1988q1
-2010q2
Offi
cia
lin
tere
stra
teA
ust
err
eic
his
che
Nati
onalb
ank
(OeN
B)
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
Ost
err
eic
his
ches
Inst
itute
for
Wir
tschaft
sfors
chung
1988q1
-2010q4
Belg
ium
GD
Pat
curr
ent
pri
ces
Banque
Nati
onale
de
Belg
ique
1995q1
-2010q2
Govern
ment
finalconsu
mpti
on
expendit
ure
Banque
Nati
onale
de
Belg
ique
1995q1
-2010q2
Pri
vate
finalconsu
mpti
on
expendit
ure
Banque
Nati
onale
de
Belg
ique
1995q1
-2010q2
GD
Pdeflato
r(2
005=
100)
Banque
Nati
onale
de
Belg
ique
1988q1
-2010q4
Offi
cia
lin
tere
stra
teIM
F1988q1
-1998q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
.
Bolivia
GD
Pat
curr
ent
pri
ces
Inst
ituto
Nacio
nalde
Est
adıs
tica/B
anco
Centr
alde
Bolivia
1993q2
-2010q2
Public
consu
mpti
on
Inst
ituto
Nacio
nalde
Est
adıs
tica/B
anco
Centr
alde
Bolivia
1993q2
-2010q2
Pri
vate
consu
mpti
on
Inst
ituto
Nacio
nalde
Est
adıs
tica/B
anco
Centr
alde
Bolivia
1993q2
-2010q2
GD
PIm
plicit
Deflato
r(1
990=
100)
Inst
ituto
Nacio
nalde
Est
adıs
tica/B
anco
Centr
alde
Bolivia
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)W
eekly
Bank
Report
s.B
anco
Centr
alde
Bolivia
,IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
.
Bra
zil
GD
Pat
curr
ent
pri
ces
Fundacao
Inst
ituto
Bra
sile
iro
de
Geogra
fia
eEst
ati
stic
a.
1990q2
-2010q2
Govern
ment
consu
mpti
on
Fundacao
Inst
ituto
Bra
sile
iro
de
Geogra
fia
eEst
ati
stic
a.
1994q1
-2010q2
Pri
vate
consu
mpti
on
Fundacao
Inst
ituto
Bra
sile
iro
de
Geogra
fia
eEst
ati
stic
a.
1994q1
-2010q2
Whole
sale
Pri
ce
Index,dom
est
icsu
pply
(1994=
100)
Fundacao
Getu
lio
Varg
as
SELIC
(targ
et
rate
)B
anco
Centr
aldo
Bra
sil,
IMF
1998q1
-2010q4
44
–conti
nued
from
pre
vio
us
page
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
(econom
ically
acti
ve
popula
tion)
Fundacao
Inst
ituto
Bra
sile
iro
de
Geogra
fia
eEst
ati
stic
a.
2001q3
-2010q3
Canadı
GD
Pat
curr
ent
pri
ces
Sta
tist
ics
Canada
1988q1
-2010q2
Govern
ment
consu
mpti
on
Sta
tist
ics
Canada
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
sSta
tist
ics
Canada
1988q1
-2010q2
GD
Pdeflato
r(2
002=
100)
Sta
tist
ics
Canada
1988q1
-2010q2
Targ
et
for
the
overn
ight
rate
Bank
ofC
anada
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent:
15
years
&over
(thous)
Sta
tist
ics
Canada
1988q1
-2010q3
Chile
GD
Pat
curr
ent
pri
ces
Banco
Centr
alde
Chile
1990q1
-2010q2
Govern
ment
consu
mpti
on
Banco
Centr
alde
Chile
1996q1
-2010q2
Consu
mpti
on
Banco
Centr
alde
Chile
1996q1
-2010q2
GD
Pim
plicit
deflato
r(2
003=
100)
Banco
Centr
alde
Chile
1988q1
-2010q4
Moneta
ryPolicy
rate
Banco
Centr
alde
Chile,IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
Nati
onalB
ure
au
ofSta
tist
ics
1988q1
-2010q4
Chin
aG
DP
at
curr
ent
pri
ces
Chin
aN
ati
onalB
ure
au
ofSta
tist
ics.
1992q1
-2010q3
Public
consu
mpti
on
Chin
aN
ati
onalB
ure
au
ofSta
tist
ics.
1988q1
-2009q4
Resi
denti
alconsu
mpti
on
Chin
aN
ati
onalB
ure
au
ofSta
tist
ics.
1988q1
-2009q4
GD
Pdeflato
r(1
990=
100)
Chin
aN
ati
onalB
ure
au
ofSta
tist
ics.
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1992q2
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
Sta
teSta
tist
icalO
ffice
1988q1
-2007q4
Colo
mbia
GD
Pat
curr
ent
pri
ces
Depart
am
ento
Adm
inis
trati
vo
Nacio
nalde
Est
adıs
tica
(DA
NE)
1994q1
-2009q4
Govern
ment
consu
mpti
on
Depart
am
ento
Adm
inis
trati
vo
Nacio
nalde
Est
adıs
tica
(DA
NE)
1994q1
-2009q4
Pri
vate
consu
mpti
on
Depart
am
ento
Adm
inis
trati
vo
Nacio
nalde
Est
adıs
tica
(DA
NE)
2000q1
-2010q2
GD
Pim
plicit
deflato
r(2
000=
100)
Depart
am
ento
Adm
inis
trati
vo
Nacio
nalde
Est
adıs
tica
(DA
NE)
1988q1
-2010q2
Inte
rventi
on
rate
Banco
de
laR
epublica
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
yed
work
ing
age
popula
tion
Depart
am
ento
Adm
inis
trati
vo
Nacio
nalde
Est
adıs
tica
(DA
NE)
2001q1
-2010q4
Cost
aR
ica
GD
Pat
curr
ent
pri
ces
Banco
Centr
alde
Cost
aR
ica
1991q1
-2010q2
Govern
ment
consu
mpti
on
Banco
Centr
alde
Cost
aR
ica.
1994q1
-2010q2
Pri
vate
consu
mpti
on
Banco
Centr
alde
Cost
aR
ica
1994q1
-2010q2
GD
PIm
plicit
Deflato
r(1
991=
100)
Banco
Centr
alde
Cost
aR
ica.
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
.
Czech
Republic
GD
Pat
curr
ent
pri
ces
Czech
Sta
tist
icalO
ffice
1995q1
-2010q2
Govern
ment
consu
mpti
on
Czech
Sta
tist
icalO
ffice
1995q1
-2010q2
Pri
vate
consu
mpti
on
Czech
Sta
tist
icalO
ffice
1995q1
-2010q2
45
–conti
nued
from
pre
vio
us
page
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
GD
PIm
plicit
Deflato
rC
zech
Sta
tist
icalO
ffice
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1993q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1990q1
-2010q4
Em
plo
ym
ent
NA
.
Denm
ark
GD
Pat
curr
ent
pri
ces
Danm
ark
sSta
tist
ik1990q1
-2010q2
Govern
ment
consu
mpti
on
Danm
ark
sSta
tist
ik1990q1
-2010q2
Pri
vate
consu
mpti
on
Danm
ark
sSta
tist
ik1990q1
-2010q2
GD
PIm
plicit
Deflato
r(1
991=
100)
Danm
ark
sSta
tist
ik1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
.
Ecuador
GD
Pat
curr
ent
pri
ces
Banco
Centr
aldelEcuador
1990q1
-2010q2
Govern
ment
consu
mpti
on
Banco
Centr
aldelEcuador
1990q1
-2010q2
Pri
vate
consu
mpti
on
Banco
Centr
aldelEcuador
1990q1
-2010q2
GD
Pdeflato
rB
anco
Centr
aldelEcuador
1988q1
-2010q2
Basi
ccentr
alB
ank
Rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent:
glo
baloccupati
on
rate
(%)
Banco
Centr
aldelEcuador
1998q3
-2008q4
ElSalv
ador
GD
Pat
curr
ent
pri
ces
Banco
Centr
alde
Rese
rva
delSalv
ador
1990q1
-2010q2
Public
consu
mpti
on
Banco
Centr
alde
Rese
rva
delSalv
ador
1993q2
-2009q4
Pri
vate
consu
mpti
on
Banco
Centr
alde
Rese
rva
delSalv
ador
1993q2
-2009q4
GD
PD
eflato
rH
aver
Analy
tics
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1995q1
-2010q4
Realeffecti
ve
exchange
rate
NA
Em
plo
ym
ent
NA
.
Fin
land
GD
Pat
curr
ent
pri
ces
Tilast
okesk
us
(Sta
tist
ics
Fin
land)
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
.Pri
vate
Tilast
okesk
us
(Sta
tist
ics
Fin
land)
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
.G
overn
ment
Tilast
okesk
us
(Sta
tist
ics
Fin
land)
1988q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Tilast
okesk
us
(Sta
tist
ics
Fin
land)
1988q1
-2010q2
Offi
cia
lin
tere
stra
te(E
nd
ofPeri
od)
Bank
ofFin
land
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
Fra
nce
GD
Pat
curr
ent
pri
ces
Inst
itut
Nati
onalde
laSta
tist
ique
et
des
Etu
des
Econom
iques.
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Public
Inst
itut
Nati
onalde
laSta
tist
ique
et
des
Etu
des
Econom
iques.
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Pri
vate
Inst
itut
Nati
onalde
laSta
tist
ique
et
des
Etu
des
Econom
iques.
1988q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Inst
itut
Nati
onalde
laSta
tist
ique
et
des
Etu
des
Econom
iques.
1988q1
-2010q2
Offi
cia
lin
tere
stra
tes
Banque
de
Fra
nce
1988q1
-2010q2
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q2
Em
plo
ym
ent
Inst
itut
Nati
onalde
laSta
tist
ique
et
des
Etu
des
Econom
iques.
1995q1
-2010q2
Germ
any
GD
Pat
curr
ent
pri
ces
Deuts
che
Bundesb
ank
1988q1
-2010q2
46
–conti
nued
from
pre
vio
us
page
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
Fin
alconsu
mpti
on
expendit
ure
s.Public
Deuts
che
Bundesb
ank
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Pri
vate
Deuts
che
Bundesb
ank
1988q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Deuts
che
Bundesb
ank
1988q1
-2010q2
Offi
cia
lin
tere
stra
tes
Deuts
che
Bundesb
ank
1988q1
-2010q2
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q2
Em
plo
ym
ent
(Thousa
nds)
Deuts
che
Bundesb
ank
1988q1
-2010q2
Gre
ece
GD
Pat
curr
ent
pri
ces
Nati
onalSta
tist
icalServ
ice
ofG
reece
2000q1
-2010q2
Public
consu
mpti
on
Nati
onalSta
tist
icalServ
ice
ofG
reece
2000q1
-2010q2
Pri
vate
consu
mpti
on
Nati
onalSta
tist
icalServ
ice
ofG
reece
2000q1
-2010q2
GD
Pdeflato
r(2
000=
100).
Quart
erl
yN
ati
onalSta
tist
icalServ
ice
ofG
reece
2000q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2000q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
Guate
mala
GD
Pat
curr
ent
pri
ces
Banco
de
Guate
mala
2001q1
-2009q4
Govern
ment
consu
mpti
on
Banco
de
Guate
mala
2001q1
-2009q4
Pri
vate
consu
mpti
on
Banco
de
Guate
mala
2001q1
-2009q4
GD
Pdeflato
rB
anco
de
Guate
mala
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-1992q2
Realeffecti
ve
exchange
rate
NA
Em
plo
ym
ent
NA
Hong
Kong
GD
Pat
curr
ent
pri
ces
HK
Censu
sand
Sta
tist
ics
Depart
ment
1988q1
-2010q3
Govern
ment
consu
mpti
on
expendit
ure
HK
Censu
sand
Sta
tist
ics
Depart
ment
1988q1
-2010q3
Pri
vate
consu
mpti
on
expendit
ure
HK
Censu
sand
Sta
tist
ics
Depart
ment
1988q1
-2010q3
GD
Pdeflato
r(2
000=
100)
HK
Censu
sand
Sta
tist
ics
Depart
ment
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
Hungary
GD
Pat
curr
ent
pri
ces
Centr
alSta
tist
icalO
ffice
1995q1
-2010q2
Govern
ment
consu
mpti
on
expendit
ure
Centr
alSta
tist
icalO
ffice
1995q1
-2010q2
Pri
vate
consu
mpti
on
expendit
ure
Centr
alSta
tist
icalO
ffice
1995q1
-2010q2
GD
Pdeflato
rC
entr
alSta
tist
icalO
ffice
(Haver
Analy
tics)
1995q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1990q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
Icela
nd
GD
Pat
curr
ent
pri
ces
Sta
tist
ics
Icela
nd
1997q1
-2010q2
Govern
ment
finalconsu
mpti
on
Sta
tist
ics
Icela
nd
1997q1
-2010q2
Pri
vate
finalconsu
mpti
on
Sta
tist
ics
Icela
nd
1997q1
-2010q2
GD
Pim
plicit
pri
ce
deflato
r(2
000=
100)
Sta
tist
ics
Icela
nd
1988q1
-2010q2
Inte
rbank
Offere
dR
ate
(REIB
OR
)C
entr
alB
ank
ofIc
ela
nd
1994q2
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
47
–conti
nued
from
pre
vio
us
page
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
Em
plo
ym
ent
NA
.
India
GD
Pat
curr
ent
pri
ces
Centr
alSta
tist
icalO
rganiz
ati
on,In
dia
.1996q2
-2010q2
Govern
ment
spendin
gIS
FD
ATA
1996q2
-2010q2
Pri
vate
consu
mpti
on
expendit
ure
Centr
alSta
tist
icalO
rganiz
ati
on,In
dia
.1996q2
-2010q2
GD
Pdeflato
r(F
iscalYear
1999=
100)
Centr
alSta
tist
icalO
rganiz
ati
on,In
dia
.1988q1
-2010q4
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Tota
lem
plo
ym
ent
Rese
rve
Bank
ofIn
dia
1988q1
-2007q2
Indonesi
aG
DP
at
curr
ent
pri
ces
Bir
oPusa
tSta
tist
ic1988q1
-2010q3
Genera
lgovern
ment
consu
mpti
on
expendit
ure
Bir
oPusa
tSta
tist
ic2000q1
-2010q3
Pri
vate
consu
mpti
on
expendit
ure
Bir
oPusa
tSta
tist
ic2000q1
-2010q3
GD
Pdeflato
r(2
000=
100)
Bir
oPusa
tSta
tist
ic1988q1
-2010q3
Dis
count
rate
(policy
inte
rest
rate
)IM
F1990q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1994q1
-2010q4
Em
plo
ym
ent
NA
.
Irela
nd
GD
Pat
curr
ent
pri
ces
&exchange
rate
s(M
il.U
S$)
Centr
alSta
tist
ics
Offi
ce
1988q1
-2010q2
Govern
ment
consu
mpti
on
expendit
ure
Centr
alSta
tist
ics
Offi
ce
1997q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Pri
vate
Centr
alSta
tist
ics
Offi
ce
1997q1
-2010q2
GD
Pdeflato
r(2
005=
100)
Centr
alSta
tist
ics
Offi
ce
1988q1
-2010q4
EC
Bm
ain
refinancin
gra
te:
min
imum
bid
rate
Centr
alSta
tist
ics
Offi
ce
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
.
Italy
GD
Pat
curr
ent
pri
ces
Isti
tuto
Nazio
nale
diSta
tist
ica
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Public
Isti
tuto
Nazio
nale
diSta
tist
ica
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Pri
vate
Isti
tuto
Nazio
nale
diSta
tist
ica
1988q1
-2010q2
GD
Pdeflato
r(S
A,2000=
100
Isti
tuto
Nazio
nale
diSta
tist
ica
1988q1
-2010q2
Offi
cia
lin
tere
stra
te(E
nd
ofPeri
od)
Banca
d’Ita
lia
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
Isti
tuto
Nazio
nale
diSta
tist
ica
1988q1
-2010q2
Japan
GD
Pat
curr
ent
pri
ces
Cabin
et
Offi
ce
1988q1
-2010q2
Govern
ment
consu
mpti
on
expendit
ure
Cabin
et
Offi
ce
1988q1
-2010q2
Pri
vate
consu
mpti
on
expendit
ure
sC
abin
et
Offi
ce
1988q1
-2010q2
GD
Pdeflato
rC
abin
et
Offi
ce
1988q1
-2010q2
Overn
ight
call
rate
:uncollate
ralized
(targ
et)
Bank
ofJapan
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
.
Kore
aG
DP
at
curr
ent
pri
ces
The
Bank
ofK
ore
a1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
.G
overn
ment
The
Bank
ofK
ore
a1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
.Pri
vate
The
Bank
ofK
ore
a1988q1
-2010q2
GD
Pdeflato
r(2
000=
100)
The
Bank
ofK
ore
a1988q1
-2010q2
48
–conti
nued
from
pre
vio
us
page
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Tota
lem
plo
ym
ent
Nati
onalSta
tist
icalO
ffice
1988q1
-2010q3
Luxem
bourg
GD
Pat
curr
ent
pri
ces
Luxem
bourg
Centr
alServ
ice
ofSta
tist
ics
and
Econom
icStu
die
s.1995q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Public
Luxem
bourg
Centr
alServ
ice
ofSta
tist
ics
and
Econom
icStu
die
s.1995q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Pri
vate
Luxem
bourg
Centr
alServ
ice
ofSta
tist
ics
and
Econom
icStu
die
s.1995q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Luxem
bourg
Centr
alServ
ice
ofSta
tist
ics
and
Econom
icStu
die
s.1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1990q1
-1999q1
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
.
Mala
ysi
aG
DP
at
curr
ent
pri
ces
Depart
ment
ofSta
tist
ics
1991q1
-2010q2
Govern
ment
finalconsu
mpti
on
expendit
ure
Depart
ment
ofSta
tist
ics
1991q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
.Pri
vate
Depart
ment
ofSta
tist
ics
1991q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Depart
ment
ofSta
tist
ics
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F2004q2
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
yed
Depart
ment
ofSta
tist
ics
1998q1
-2010q2
Mexic
oG
DP
at
curr
ent
pri
ces
Inst
ituto
Nacio
nalde
Est
adis
tica
Geogra
fia
eIn
form
ati
ca
1988q1
-2010q2
Govern
ment
consu
mpti
on
ISFT
DATA
1993q1
-2010q2
Pri
vate
consu
mpti
on
Inst
ituto
Nacio
nalde
Est
adis
tica
Geogra
fIa
eIn
form
ati
ca
1993q1
-2010q2
GD
Pdeflato
r(1
993=
100)
Inst
ituto
Nacio
nalde
Est
adis
tica
Geogra
fIa
eIn
form
ati
ca
1988q1
-2010q2
Targ
et
rate
Banco
de
Mexic
o,IM
F1995q2
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
(pers
ons)
Inst
ituto
Nacio
nalde
Est
adis
tica
Geogra
fIa
eIn
form
ati
ca
2005q1
-2010q3
Neth
erl
ands
GD
Pat
curr
ent
pri
ces
Centr
aalbure
au
voor
de
stati
stie
k1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Public
Centr
aalbure
au
voor
de
stati
stie
k1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Pri
vate
Centr
aalbure
au
voor
de
stati
stie
k1988q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Centr
aalbure
au
voor
de
stati
stie
k1988q1
-2010q2
EC
Bm
ain
refinancin
gra
te:
min
imum
bid
rate
De
Nederl
andsc
he
Bank
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Tota
lem
plo
ym
ent
NA
New
Zeala
nd
GD
Pat
curr
ent
pri
ces
Sta
tist
ics
New
Zeala
nd
1988q1
-2010q2
Public
consu
mpti
on
expendit
ure
Sta
tist
ics
New
Zeala
nd
1988q1
-2010q2
Pri
vate
consu
mpti
on
expendit
ure
Sta
tist
ics
New
Zeala
nd
1988q1
-2010q2
GD
Pdeflato
r(S
A,Q
31995-Q
21996=
1000)
Sta
tist
ics
New
Zeala
nd
1988q1
-2010q2
Offi
cia
lcash
rate
(EO
P)
Rese
rve
Bank
ofN
ew
Zeala
nd
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Tota
lem
plo
ym
ent
NA
Norw
ay
GD
Pat
curr
ent
pri
ces
Sta
tist
isk
Sentr
alb
yra
1988q1
-2010q2
Genera
lgovern
ment
consu
mpti
on
Sta
tist
isk
Sentr
alb
yra
1988q1
-2010q2
49
–conti
nued
from
pre
vio
us
page
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
Pri
vate
consu
mpti
on
Sta
tist
isk
Sentr
alb
yra
1988q1
-2010q2
GD
Pdeflato
r(2
005=
100)
Sta
tist
isk
Sentr
alb
yra
1988q1
-2010q2
Sig
ht
deposi
tra
teN
org
es
Bank
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Tota
lem
plo
ym
ent
NA
Pakis
tan
GD
Pat
curr
ent
pri
ces
Federa
lB
ure
au
ofSta
tist
ics
1988q4
-2010q2
Genera
lgovern
ment
Federa
lB
ure
au
ofSta
tist
ics
1998q3
-2010q2
Tota
lconsu
mpti
on.
Pri
vate
Federa
lB
ure
au
ofSta
tist
ics
1998q3
-2010q2
GD
Pdeflato
r(F
iscalyear
2000=
100)
Federa
lB
ure
au
ofSta
tist
ics
1988q1
-2010q4
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
(10
Years
and
over)
Federa
lB
ure
au
ofSta
tist
ics
1989q3
-2009q2
Para
guay
GD
Pat
curr
ent
pri
ces
Banco
Centr
aldelPara
guay
1994q1
-2010q2
Govern
ment
consu
mpti
on
Banco
Centr
aldelPara
guay
1994q1
-2010q2
Pri
vate
consu
mpti
on
Banco
Centr
aldelPara
guay
1994q1
-2010q2
GD
PIm
plicit
Deflato
r(1
982=
100)
Banco
Centr
aldelPara
guay
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1989q4
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
.
Peru
GD
Pat
curr
ent
pri
ces
Min
iste
rio
de
Econom
iay
Obra
sy
Serv
icio
sPublicos
(Haver
analy
tics)
1989q1
-2010q2
Public
consu
mpti
on
Min
iste
rio
de
Econom
iay
Obra
sy
Serv
icio
sPublicos
(Haver
analy
tics)
1992q3
-2010q2
Pri
vate
consu
mpti
on
Min
iste
rio
de
Econom
iay
Obra
sy
Serv
icio
sPublicos
(Haver
analy
tics)
1992q3
-2010q2
GD
PIm
plicit
Deflato
r(1
994=
100)
Min
iste
rio
de
Econom
iay
Obra
sy
Serv
icio
sPublicos
(Haver
analy
tics)
1988q1
-2010q2
Refe
rence
rate
(policy
inte
rest
rate
)IM
F1991q3
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1994q1
-2010q4
Em
plo
yed:
Metr
opolita
nLim
aB
anco
Centr
alde
Rese
rva
delPeru
2001q2
-2010q3
Philip
pin
es
GD
Pat
curr
ent
pri
ces
Nati
onalEconom
icand
Develo
pm
ent
Auth
ori
ty1988q1
-2010q2
Govern
ment
consu
mpti
on
expendit
ure
Nati
onalEconom
icand
Develo
pm
ent
Auth
ori
ty1991q3
-2010q2
Pri
vate
consu
mpti
on
Nati
onalEconom
icand
Develo
pm
ent
Auth
ori
ty1991q3
-2010q2
Gro
ssdom
est
icpro
duct
deflato
r(1
985=
100)
Nati
onalEconom
icand
Develo
pm
ent
Auth
ori
ty1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
yed
Nati
onalSta
tist
ics
Offi
ce
1990q4
-2010q3
Pola
nd
GD
Pat
curr
ent
pri
ces
Centr
alSta
tist
icalO
ffice
1995q1
-2010q2
Govern
ment
consu
mpti
on
Centr
alSta
tist
icalO
ffice
1995q1
-2010q2
Pri
vate
consu
mpti
on
Centr
alSta
tist
icalO
ffice
1995q1
-2010q2
GD
Pdeflato
rC
entr
alSta
tist
icalO
ffice
(Haver
Analy
tics)
1995q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1990q2
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
Centr
alSta
tist
icalO
ffice
1992q2
-2010q2
50
–conti
nued
from
pre
vio
us
page
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
Port
ugal
GD
Pat
curr
ent
pri
ces
Inst
ituto
Nacio
nalde
Est
ati
stic
a1995q1
-2010q2
Public
consu
mpti
on
expendit
ure
sIn
stit
uto
Nacio
nalde
Est
ati
stic
a1995q1
-2010q2
Pri
vate
consu
mpti
on
expendit
ure
sIn
stit
uto
Nacio
nalde
Est
ati
stic
a1995q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Inst
ituto
Nacio
nalde
Est
ati
stic
a1995q1
-2010q2
EC
Bm
ain
refinancin
gra
te:
min
imum
bid
rate
Euro
pean
Centr
alB
ank
1992q1
-2010q4
Realeffecti
ve
exchange
rate
1988q1
-2010q4
Em
plo
yed
(Thousa
nd)
Inst
ituto
Nacio
nalde
Est
ati
stic
aN
A
Russ
iaG
DP
at
curr
ent
pri
ces
Sta
teC
om
itee
ofth
eR
uss
ian
Federa
tion
2003q1
-2010q2
Govern
ment
consu
mpti
on
Sta
teC
om
itee
ofth
eR
uss
ian
Federa
tion
2003q1
-2010q2
Pri
vate
consu
mpti
on
Sta
teC
om
itee
ofth
eR
uss
ian
Federa
tion
2003q1
-2010q2
GD
Pdeflato
rSta
teC
om
itee
ofth
eR
uss
ian
Federa
tion(H
aver
Analy
tics)
2003q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1996q3
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
Sta
teC
om
itee
ofth
eR
uss
ian
Federa
tion
1991q1
-2010q3
Sin
gapore
GD
Pat
curr
ent
pri
ces
Depart
ment
ofSta
tist
ics
1988q1
-2010q2
Govern
ment
consu
mpti
on
Depart
ment
ofSta
tist
ics
1988q1
-2010q2
Pri
vate
consu
mpti
on
Depart
ment
ofSta
tist
ics
1988q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Depart
ment
ofSta
tist
ics
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
Slo
vakia
GD
Pat
curr
ent
pri
ces
Sta
tist
icalO
ffice
ofth
eSlo
vak
Republic
1995q1
-2010q2
Govern
ment
consu
mpti
on
Sta
tist
icalO
ffice
ofth
eSlo
vak
Republic
1995q1
-2010q2
Pri
vate
consu
mpti
on
Sta
tist
icalO
ffice
ofth
eSlo
vak
Republic
1995q1
-2010q2
GD
Pdeflato
rSta
tist
icalO
ffice
ofth
eSlo
vak
Republic
(Haver
Analy
tics)
1995q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1995q3
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1990q1
-2010q2
Em
plo
ym
ent
NA
South
Afr
ica
GD
Pat
curr
ent
pri
ces
South
Afr
ican
Rese
rve
Bank
1988q1
-2010q2
Govern
ment
consu
mpti
on
South
Afr
ican
Rese
rve
Bank
1988q1
-2010q2
Pri
vate
consu
mpti
on
South
Afr
ican
Rese
rve
Bank
1988q1
-2010q2
GD
Pdeflato
rSouth
Afr
ican
Rese
rve
Bank
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
South
Afr
ican
Rese
rve
Bank
2000q1
-2010q2
Spain
GD
Pat
curr
ent
pri
ces
Inst
ituto
Nacio
nalde
Est
adis
tica
1988q1
-2010q2
Genera
lgovern
ment
consu
mpti
on
exp
Inst
ituto
Nacio
nalde
Est
adis
tica
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
Inst
ituto
Nacio
nalde
Est
adis
tica
1988q1
-2010q2
GD
Pdeflato
r(2
000=
100)(
Q1
2000-p
rese
nt)
Inst
ituto
Nacio
nalde
Est
adis
tica
1988q1
-2010q2
Offi
cia
lin
tere
stra
teEuro
pean
Centr
alB
ank
1988q1
-2010q4
51
–conti
nued
from
pre
vio
us
page
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
(Thous)
Inst
ituto
Nacio
nalde
Est
adis
tica
2000q1
-2010q3
Sw
eden
GD
Pat
curr
ent
pri
ces
Sta
tist
iska
Centr
alb
yra
n(S
tati
stic
sSw
eden)
1993q1
-2010q2
Govern
ment
consu
mpti
on
Sta
tist
iska
Centr
alb
yra
n(S
tati
stic
sSw
eden)
1993q1
-2010q2
Pri
vate
consu
mpti
on
expendit
ure
sSta
tist
iska
Centr
alb
yra
n(S
tati
stic
sSw
eden)
1993q1
-2010q2
Quart
erl
yG
DP
deflato
r(2
000=
100)
Sta
tist
iska
Centr
alb
yra
n(S
tati
stic
sSw
eden)
1988q1
-2010q2
Rik
sbank
repo
rate
Sveri
ges
Rik
sbank
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
Sw
itzerl
and
GD
Pat
curr
ent
pri
ces
Sta
ats
sekre
tari
at
fır
Wir
tschaft
1993q1
-2010q2
Public
consu
mpti
on
expendit
ure
Sta
ats
sekre
tari
at
fır
Wir
tschaft
1993q1
-2010q2
Pri
vate
consu
mpti
on
Sta
ats
sekre
tari
at
fır
Wir
tschaft
1993q1
-2010q2
Quart
erl
yG
DP
Deflato
r(2
000=
100)
Sta
ats
sekre
tari
at
fır
Wir
tschaft
1988q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
Taiw
an
GD
Pat
curr
ent
pri
ces
Dir
ecto
rate
-Genera
lofB
udget,
Accounti
ng
&Sta
tist
ics,
Executi
ve
Yuan
(DG
BA
SY
).1988q1
-2010q2
Govern
ment
finalconsu
mpti
on
expendit
ure
Dir
ecto
rate
-Genera
lofB
udget,
Accounti
ng
&Sta
tist
ics,
Executi
ve
Yuan
(DG
BA
SY
).1988q1
-2010q2
Pri
vate
finalconsu
mpti
on
expendit
ure
Dir
ecto
rate
-Genera
lofB
udget,
Accounti
ng
&Sta
tist
ics,
Executi
ve
Yuan
(DG
BA
SY
).1988q1
-2010q2
GD
Pdeflato
r(2
001=
100)
Dir
ecto
rate
-Genera
lofB
udget,
Accounti
ng
&Sta
tist
ics,
Executi
ve
Yuan
(DG
BA
SY
).1988q1
-2010q2
Redis
count
rate
Centr
alB
ank
ofC
hin
a1988q1
-2010q4
Realeffecti
ve
exchange
rate
NA
Em
plo
ym
ent
NA
Thailand
GD
Pat
curr
ent
pri
ces
Nati
onalEconom
icand
Socia
lD
evelo
pm
ent
Board
(NESB
D)
1993q1
-2010q2
Govern
ment
consu
mpti
on
expendit
ure
Nati
onalEconom
icand
Socia
lD
evelo
pm
ent
Board
(NESD
B)
1993q1
-2010q2
Pri
vate
consu
mpti
on
expendit
ure
Nati
onalEconom
icand
Socia
lD
evelo
pm
ent
Board
(NESD
B)
1993q1
-2010q2
GD
Pdeflato
r(1
988=
100)
Nati
onalEconom
icand
Socia
lD
evelo
pm
ent
Board
(NESD
B)
1993q1
-2010q2
Policy
targ
et
rate
(%,EO
P)
Bank
ofT
hailand,IM
F1988q1
2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1994q1
-2010q4
Em
plo
ym
ent
NA
Turk
ey
GD
Pat
curr
ent
pri
ces
Haver
Analy
tics
1988q1
-2010q2
Govern
ment
finalconsu
mpti
on
expendit
ure
Haver
Analy
tics
1993q1
-2010q2
Pri
vate
finalconsu
mpti
on
expendit
ure
Haver
Analy
tics
1989q4
-2010q2
GD
Pdeflato
rH
aver
Analy
tics
1988q1
2010q2
Redis
count
rate
IMF
1988q1
-2010q2
Realeffecti
ve
exchange
rate
IMF,B
IS1994q1
-2010q4
Em
plo
ym
ent
Haver
Analy
tics
1991q3
-2010q2
Unit
ed
Kin
gdom
GD
Pat
curr
ent
pri
ces
Offi
ce
for
Nati
onalSta
tist
ics
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Public
Offi
ce
for
Nati
onalSta
tist
ics
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Pri
vate
Offi
ce
for
Nati
onalSta
tist
ics
1988q1
-2010q2
52
–conti
nued
from
pre
vio
us
page
CO
UN
TRY
VA
RIA
BLE
NA
ME
SO
UR
CE
TIM
EPER
IOD
GD
Pdeflato
r(2
003=
100)
Offi
ce
for
Nati
onalSta
tist
ics
1988q1
-2010q2
Base
rate
(repo
rate
),B
ank
ofEngla
nd
(EO
P)
Bank
ofEngla
nd
1988q1
-2010q4
Gro
ssfixed
capit
alfo
rmati
on
Offi
ce
for
Nati
onalSta
tist
ics
1988q1
-2010q4
Em
plo
ym
ent
(Thous)
Offi
ce
for
Nati
onalSta
tist
ics
1988q1
-2010q2
Unit
ed
Sta
tes
Gro
ssdom
est
icpro
duct
Bure
au
ofEconom
icA
naly
sis
1988q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Public
Bure
au
ofEconom
icA
naly
sis
1988q1
-2010q2
Pri
vate
(pers
onalconsu
mpti
on
expendit
ure
)B
ure
au
ofEconom
icA
naly
sis
1988q1
-2010q2
GD
Pdeflato
r(2
000=
100)
Bure
au
ofEconom
icA
naly
sis
1988q1
-2010q2
Federa
lFunds
Targ
et
Rate
U.S
.Federa
lR
ese
rve
Bank
1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
U.S
.B
ure
au
ofLabor
Sta
tist
ics
1988q1
-2010q2
Venezuela
GD
Pat
curr
ent
pri
ces
Banco
Centr
alde
Venezuela
1997q1
-2009q4
Govern
ment
consu
mpti
on
Banco
Centr
alde
Venezuela
1998q1
-2009q4
Pri
vate
consu
mpti
on
Banco
Centr
alde
Venezuela
1998q1
-2009q4
GD
Pim
plicit
deflato
r(1
997=
100)
Banco
Centr
alde
Venezuela
1998q1
-2009q4
Dis
count
rate
(policy
inte
rest
rate
)IM
F1988q1
-2010q4
Realeffecti
ve
exchange
rate
IMF,B
IS1988q1
-2010q4
Em
plo
ym
ent
NA
Vie
tnam
GD
Pat
curr
ent
pri
ces
Genera
lSta
tist
ics
Offi
ce
ofV
ietn
am
1993q2
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Public
Genera
lSta
tist
ics
Offi
ce
ofV
ietn
am
1995q1
-2010q2
Fin
alconsu
mpti
on
expendit
ure
s.Pri
vate
Genera
lSta
tist
ics
Offi
ce
ofV
ietn
am
1995q1
-2010q2
GD
Pdeflato
r(1
994=
100)
Genera
lSta
tist
ics
Offi
ce
ofV
ietn
am
1995q1
-2010q2
Dis
count
rate
(policy
inte
rest
rate
)IM
F1996q1
-2010q4
Realeffecti
ve
exchange
rate
NA
Em
plo
ym
ent
NA
53